Period Count for Fibonacci recurrences modulo primes

Let pp be prime, and let #(p)\#(p) count distinct periods modulo pp of the Fibonacci recurrence and its parity transform over all initial conditions (a0,a1)∈(Z/pZ)2(a_0,a_1)\in(\mathbb Z/p\mathbb Z)^2. Let α\alpha be the positive integer governing the Pisano period. Period Count for Fibonacci Recurrences modulo pp. If p≡2,3(mod5)p\equiv2,3\pmod5, then πA(p)=2(p+1)/α\pi_A(p)=2(p+1)/\alpha for odd α\alpha and

#A(p)=α2(p−1)+1.\#_A(p)=\frac\alpha2(p-1)+1.

If p≡1,4(mod5)p\equiv1,4\pmod5, then πB(p)=(p−1)/α\pi_B(p)=(p-1)/\alpha, with subclasses

#B1(p)=α(p+1)+1,\#_{B1}(p)=\alpha(p+1)+1,

for the two-length case, and

#B2(p)=α(p+2)+1,\#_{B2}(p)=\alpha(p+2)+1,

for the three-length case. In subclass B2, the lengths are 00, πB(p)/2\pi_B(p)/2, and πB(p)\pi_B(p), with multiplicities 2α2\alpha and pαp\alpha for the latter two. All primes congruent to 11,19(mod20)11,19\pmod{20} belong to B2, while primes congruent to 1,9(mod20)1,9\pmod{20} may belong to either subclass. These proposed classifications and counts are not resolved in the supplied text.

References

Primary source

Marc T. Pudelko, “Modular Periodicity of Random Initialized Recurrences”, arXiv:2510.24882 (2026).

Progress summary

Refreshed
Claimed progress

The formulas remain conjectural, while an unverified submitted proof claims to establish them and no independent confirmation was found.

The conjecture, posed by Marc Thomas Pudelko, classifies all periods arising from the Fibonacci recurrence and its parity transform over initial states modulo a prime. The October 2025 preprint states the formulas as Conjecture 5, not as a theorem.

Known results

  • Wall and Robinson established that the classical Fibonacci period divides p−1p-1 when p≡1,4(mod5)p\equiv1,4\pmod5 and divides 2(p+1)2(p+1) when p≡2,3(mod5)p\equiv2,3\pmod5.
  • The exceptional classical periods are 2020 for p=5p=5 and 33 for p=2p=2.
  • These divisibility results do not determine the proposed counts of distinct periods.

Community submission (unverified)

A submitted proof argues that the two transition matrices act as invertible linear maps and claims a complete orbit enumeration, including the inert-prime formula, both split-prime subclasses, and separate treatment of p=2p=2 and p=5p=5. Its correctness has not been independently verified.

Current status (as of August 2026): The period-count formulas remain unproved in the retrieved public record; the submitted proof is substantive but unverified.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

Complete prime-modulus classification of both Fibonacci recurrences

Consider the two invertible state-transition matrices

U=(0111),V=(011−1)over Fp.(1)U=\begin{pmatrix}0&1\\1&1\end{pmatrix}, \qquad V=\begin{pmatrix}0&1\\1&-1\end{pmatrix} \quad\text{over }\mathbb F_p. \tag{1}

Distinct periodic sequences up to cyclic shift correspond exactly to the orbits of these matrices on Fp2\mathbb F_p^2. The zero state contributes one orbit of length 11.

We prove every asserted orbit count and orbit-length multiplicity in Marc T. Pudelko's Conjecture 5. We also handle the exceptional primes 22 and 55, which require separate treatment. Classical root-order and rank-of-apparition results are discussed by Ballot and Elia; the argument below supplies the complete orbit enumeration requested in the conjecture.

Inert primes

Assume first that pp is odd, p≠5p\neq5, and

(5p)=−1,equivalentlyp≡2,3(mod5).(2)\left(\frac5p\right)=-1, \qquad\text{equivalently}\qquad p\equiv2,3\pmod5. \tag{2}

The characteristic polynomial

f(X)=X2−X−1(3)f(X)=X^2-X-1 \tag{3}

is irreducible over Fp\mathbb F_p. If u∈Fp2u\in\mathbb F_{p^2} is a root, its conjugate is the other root, so

up=−u−1,up+1=−1.(4)u^p=-u^{-1}, \qquad u^{p+1}=-1. \tag{4}

Write t=ord⁡(u)t=\operatorname{ord}(u). Equation (4) implies

t∣2(p+1),t∤p+1.(5)t\mid2(p+1), \qquad t\nmid p+1. \tag{5}

Consequently

t=2(p+1)α,α odd.(6)t=\frac{2(p+1)}{\alpha}, \qquad \alpha\text{ odd}. \tag{6}

Indeed, if α=2(p+1)/t\alpha=2(p+1)/t were even, then tt would divide p+1p+1, contradicting (5).

The companion-matrix identification Fp2≅Fp2\mathbb F_p^2\cong\mathbb F_{p^2} identifies UU with multiplication by uu. Every nonzero state therefore has the same orbit length:

ujz=z(z≠0)⟺uj=1.(7)u^jz=z\quad(z\neq0) \quad\Longleftrightarrow\quad u^j=1. \tag{7}

Thus π(p)=t\pi(p)=t, and the complete orbit distribution is

1 orbit of length 1,p2−1π(p)=α(p−1)2 orbits of length π(p).(8)\boxed{ 1\text{ orbit of length }1, \qquad \frac{p^2-1}{\pi(p)} =\frac{\alpha(p-1)}2 \text{ orbits of length }\pi(p). } \tag{8}

In particular,

#A(p)=1+α(p−1)2.(9)\#_A(p)=1+\frac{\alpha(p-1)}2. \tag{9}

Split primes and the two subclasses

Suppose instead that

(5p)=1,equivalentlyp≡1,4(mod5).(10)\left(\frac5p\right)=1, \qquad\text{equivalently}\qquad p\equiv1,4\pmod5. \tag{10}

Let u,v∈Fp×u,v\in\mathbb F_p^\times be the two roots of ff, and put

v=−u−1,t=ord⁡(u),t′=ord⁡(v).(11)v=-u^{-1}, \qquad t=\operatorname{ord}(u), \qquad t'=\operatorname{ord}(v). \tag{11}

The root orders satisfy

t′={2t,t odd,t/2,t≡2(mod4),t,t≡0(mod4).(12)t'= \begin{cases} 2t,&t\text{ odd},\\ t/2,&t\equiv2\pmod4,\\ t,&t\equiv0\pmod4. \end{cases} \tag{12}

For odd tt, adjoining the factor −1-1 doubles the order. For even tt, use −1=ut/2-1=u^{t/2} to obtain

v=ut/2−1,t′=tgcd⁡(t,t/2−1),(13)v=u^{t/2-1}, \qquad t'=\frac{t}{\gcd(t,t/2-1)}, \tag{13}

which gives the remaining two cases.

Diagonalizing UU, its action becomes

(x,y)⟼(ux,vy).(14)(x,y)\longmapsto(ux,vy). \tag{14}

Therefore

π(p)=lcm⁡(t,t′)=max⁡(t,t′),α=p−1π(p).(15)\pi(p)=\operatorname{lcm}(t,t')=\max(t,t'), \qquad \alpha=\frac{p-1}{\pi(p)}. \tag{15}

If t=t′=π(p)t=t'=\pi(p), every nonzero state has orbit length π(p)\pi(p), and hence

#B1(p)=1+p2−1π(p)=1+α(p+1).(16)\boxed{ \#_{B1}(p) =1+\frac{p^2-1}{\pi(p)} =1+\alpha(p+1). } \tag{16}

Otherwise {t,t′}={π(p)/2,π(p)}\{t,t'\}=\{\pi(p)/2,\pi(p)\}. The eigenline associated with the smaller order contains p−1p-1 nonzero states and therefore contributes

p−1π(p)/2=2α(17)\frac{p-1}{\pi(p)/2}=2\alpha \tag{17}

orbits of length π(p)/2\pi(p)/2. Every other nonzero state has orbit length π(p)\pi(p), giving

p2−pπ(p)=pα(18)\frac{p^2-p}{\pi(p)}=p\alpha \tag{18}

such orbits. Thus

#B2(p)=1+2α+pα=1+α(p+2).(19)\boxed{ \#_{B2}(p)=1+2\alpha+p\alpha =1+\alpha(p+2). } \tag{19}

This also proves the two requested multiplicities, not merely the total count.

For the canonical Fibonacci initial state, a zero occurs at index jj precisely when uj=vju^j=v^j. In the unequal-order case, write π(p)=2s\pi(p)=2s, where ss is odd, and let ww denote the root of order ss. The other root is −w−1-w^{-1}, so the quotient of the roots, in one order or the other, is −w2-w^2. Because ss is odd, this quotient has order 2s=π(p)2s=\pi(p). Therefore the canonical Pisano cycle contains exactly one zero, as asserted in the conjecture.

If additionally p≡3(mod4)p\equiv3\pmod4, then −1-1 is a nonsquare. Since uv=−1uv=-1, exactly one root is a square. Because p−1≡2(mod4)p-1\equiv2\pmod4, the square root has odd order and the other root has even order. Consequently their orders differ, so

p≡11,19(mod20)⟹p belongs to subclass B2.(20)p\equiv11,19\pmod{20} \quad\Longrightarrow\quad p\text{ belongs to subclass B2}. \tag{20}

The parity transform and exceptional primes

The characteristic roots of VV are −u,−v-u,-v. Since uv=−1uv=-1,

−u=v−1,−v=u−1.(21)-u=v^{-1}, \qquad -v=u^{-1}. \tag{21}

Thus the parity transform interchanges and inverts the two root orders. In the split case this preserves both eigenline lengths and all orbit multiplicities; in the inert case it preserves the common nonzero orbit length. Hence (8), (16), and (19) hold for both recurrences.

Finally, the complete exceptional distributions are

porbit lengths, each occurring once21,351,4,20.(22)\begin{array}{c|c} p&\text{orbit lengths, each occurring once}\\ \hline 2&1,3\\ 5&1,4,20. \end{array} \tag{22}

For p=2p=2, the three nonzero states form one orbit. For p=5p=5, the repeated root is 33, which has order 44. Its eigenline contributes the unique orbit of length 44, while the nontrivial Jordan part has order 55 and makes every state outside that eigenline have period 2020.

The prime p=2p=2 shows that the source's claim that α\alpha is odd requires the implicit hypothesis that pp is odd: its Pisano period is 3=2(2+1)/23=2(2+1)/2, so its corresponding α\alpha equals 22. With that necessary endpoint clarification, the full proposed prime classification and both exceptional cases are proved.